Erik Kusch, PhD Student
Department of Biology
Section for Ecoinformatics & Biodiversity
Center for Biodiversity Dynamics in a Changing World (BIOCHANGE)
Aarhus University
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 1
Usually best addressed through
multilevel models (GLMMs)!
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 2
Attempt to account for unobserved heterogeneity.
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 3
Hidden
State
Poisson
Likeli-
hood of
n
Hierarchy of predictor variables!
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 4
Probability of discrete outcomes is
subtraction of adjacent values.
Logistic
Function
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 5
Log-cumulative
probabilities
Cumulative
Probabilities
Posterior uncertainty allows for more informative
identification of intercepts because they give us mean and
uncertainty.
Do not match exactly…
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 6
Positive Parameter estimates
indicate increase in probability
Decreasing log-cumulative odds for outcomes k-below the maximum shifts probability mass towards k.
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 7
26/03/2021
[Study Group] Bayesian Statistics with the Rethinking Material 8
Dirichlet is a distribution for probability distributions. α is the
proportion of the total mass in the resulting distribution.